Multivariate, Minimum-Curvature Splines for Randomly Spaced Data.
Abstract
This paper develops a collection of multivariate splines which are applicable to randomly spaced data. The splines are constructed from a linear combination of spline basis functions having evenly spaced nodes. A pseudoinverse solution of coefficients is obtained via a singular value decomposition (SVD). Spurious oscillations are damped by zeroing out an appropriate number of the singular values. A transformation is performed prior to entering the SVD so that, whenever the solution is nonunique, the 'minimum-curvature' solution is obtained. An added feature of these splines is an option through which a periodic fit may be produced in one or more dimensions if it is so desired. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1983
- Accession Number
- ADA132319
Entities
People
- James R. Cloutier
Organizations
- Naval Oceanographic Office